An Efficient Motion Planning Method with a Lazy Demonstration Graph for Repetitive Pick-and-Place

Area of Science:

• Robotics
• Artificial Intelligence
• Motion Planning

Background:

• Robotic systems often perform repetitive tasks requiring similar motion plans.
• Workspace changes due to operations necessitate adaptive planning strategies.
• Reusing prior planning knowledge is crucial for efficiency and adaptability.

Purpose of the Study:

• To develop a novel motion planner that leverages past successful planning cases.
• To improve adaptive sampling for robot motion planning.
• To enhance the efficiency and success rate of motion planning in dynamic environments.

Main Methods:

• Proposed the Lazy Demonstration Graph (LDG) planner, integrating successful planning cases as prior knowledge.
• Utilized a Gaussian Mixture Model (GMM) for adaptive sampling, concentrating samples in task-relevant areas.
• Combined adaptive sampling with the Lazy Probabilistic Roadmap (LazyPRM) algorithm for multi-query planning and lazy collision detection.

Main Results:

• The LDG planner demonstrated improved performance in planning time and path quality compared to state-of-the-art methods.
• Achieved higher success rates in repetitive pick-and-place tasks.
• Validated through simulations and physical experiments on an AUBO-i5 robot arm.

Conclusions:

• The proposed LDG planner effectively reuses prior knowledge for adaptive motion planning.
• The method shows robustness and validity in dynamic robotic manipulation tasks.
• Offers a significant advancement for efficient and reliable robot path planning.